Multiphase Solidification - a new process towards ,. prototypesand a new data 1I1nlit,1lo1ll"'it'l"llA1l,a

byMartin.Geiger, Wilhelm Steger

Fraunhofer-Institute for Manufacturing and Automation -Matthias Greul, Manfred Sindel

Fraunhofer-Institute for Applied Materials Research IFAM

Abstract

production of metallic and ceramic parts with technologies is requested.Multiphase Jet Solidification (MJS) is a process which reveals good results to develop acommercial system due to this task. Low viscous materials (liquefied substances orpowder-binder~pastes) are extruded through an x-y-z~ontrolled jet and parts of differentmaterials e.g. stainless steel are fabricated layer by layer up to the final extension. basicprinciple of the process and the current results will be presented. A slice format wasdesigned for MJS, but it is also usable for other technologies. The isdevelopment of a general slice interface for

Introduction

Two Institutes of the Fraunhofer Gesellschaft, Institute ManufacturingEngineering and Automation in Stuttgart (IPA) and the Institute for MaterialsResearch in Bremen (IFAM) are developing an RP suitable to produce metallic andceramic parts directly. The sum of the experiences manufacturing engineering andinformation processing atIPA and those in material research at IFAM turned out to build agood team to achieve this aim. After some preliminary examinations of ideasworking-principle was developed and named "Multiphase Solidification (MJS)".working-principle will be explained the first section and the results by producing partswith a first apparatus assembled and tested at IFAM will shown. Subject the C!P>ro.r.nn

section are examinations and developments in information processing. A slice format is, asin most RP systems, base for the computation of the NC code for MJS. Aim of the currentwork is a slice format suitable for RP processes in general, whereby a first implementationwas realized and tested to run MJS. The next task is the coordination of the differentcontributions to a 'de facto' slice standard. Such a standard is necessary, but its syntax andcontents have still to discussed.

prU1CIPle of MJS

basic idea is to extrude material through a similar to other apparatus such asthe Modeler of Stratasys or the Model Maker of Sanders Prototype.techniques the production of metallic and ceramic parts is the main aimBecause of technological limits, pure materials with a high meltingliquefied with the apparatus. Additional to liquefied purepowder-binder-mixtures are used, if the melting point of the material is too high. In anycase the material (mixture) is heated above its solidification point· and deposited bylayer. The melted substance solidifies when it contact with the the

9

previous layer due to temperature, pressure decrease and heat transfer to the part and theenvironment.

For the present research a simple apparatus is used to test the general feasibility,parameters of the building process and different materials. The main components are anx-y-z-computer controlled positioning system (machine precision +/-0,01 mm, axistraversing 500*540*175 mm3), a heated chamber and a jet system. The chamber istemperature-stabilized and can be varledwithin a range of 70 degree to 220 degree Celsius.The material is supplied as powder, pellets or bars. The material flow is controlled by apiston pressing the viscous material through the jet. The first test apparatus is shown infigure 1.

Figure 1: Alpha apparatus of Multiphase Jet Solidification

Production by liquefaction of metals

Metallic parts with a low melting point are directly fabricated by deposition out of themolten phase. The first experiments were accomplished with tin-bismuth alloys (meltingtemperature up to 180 degree Celsius) which are normally used for mould making and metalspraying. The microstructure of parts shows a good bonding between the layers.

The accuracy of the parts is still limited due to the low viscosity and the surfacetension of liquid metal. Therefore, the liquid metal contracts directly after the extrusionbefore solidification. First experiments by alloying particles into the metal in order toincrease the viscosity are very promising.

The melting temperature of industrial relevant materials like metals, e.g. zinc andaluminium, is much higher than this of the tested alloys. Therefore, it is an important task toimprove the apparatus towards higher temperatures.

10

Deposition of powder-binder mixtures

To produce prototypes of materials with higqer melting points like ceramics and themost metallic alloys the use of powder-binder mixtures reveals good results. First,prototypes made of stainless steel were fabricated via a powder-binder mixture. At a certaintemperature of about 100 degree Celsius the mixture has a suitable viscosity to extrude it viaa nozzle. Because of low surface tension the material can be deposited very exactly withoutcontraction. Subsequently the green parts are processed like a metal injection moulded(MIM) part by debinding and sintering to final density. The microstructure and themechanical properties of the prototypes are comparable to MIM parts.

Figure 2: MJS parts in stainless steel

The accuracy of the parts is higher than of the liquefied metal process. The materialbehaviour of the mixture allows a greater overhang angle. But volume shrinkage of about30% due to sintering must be taken into account when generating the control data. In MIM,the mixture is.pressed with a high velocity into a mould. Powder--binder segregation causedby the different densities of metal powder and binder could often be.observed. In MJS, thisturned out to be no problem because of the low velocity of the material through the jet incomparison with MIM. A further advantage is, that the green parts can be easily finished toa very smooth surface. Figure 2 displays sintered metallic parts (average diameter is about60 mm) out of stainless steel.

During sintering, special care has to be taken to avoid distortion by reducing thefriction between the part and the support. The wall thickness of the prototypes is currentlylimited to about 15 mm due to the debinding-process. If thicker areas are necessary, it ispossible to use hatches instead of filling the entire contour. Optimization of the bindercomponents is a further task to increase the maximum wall thickness.

Experiments with other powder-binder mixtures in order to achieve titanium andcopper parts have been carried out. The process turned out to be practicable for a wide range

11

of materials. Also the production of ceramic parts like alumina and silicon carbide is indevelopment.

The alternative to the sintering is the infiltration process. The debinded metallic partcan be infiltrated by a second metal with a lower melting temperature. This metal is suckedup by the part due to a capillary effect. The benefit of this process is the fact that there is noneed to account shrinkage.

Data pr~paration for RP

For discussions and developments On data handling in RP it is important to considerthe present global activities indeveloping the common, object oriented 'Standard for theExchange ofProduct ModelData' (STEP) /1/. In the future, the data describing geometry,topology, featllres, •materials, tolerances, etc. can be used for the planning and rapidproduction of prototypes. The directlink to different technologies e.g. Virtual Reality, CAMand CAQ with onecommon.through0l.lt.data model will become possible. In a commonproject of several Fraunhofer--Institutesactivities to develop aSTEP based data handling inRP will start this year. But international standardization needs a lot of time. Until STEP isapplicable for RP, the industry/heeds fast •and suitable solutions fitting the differentapplications best. The data handling, which will be discussed here, could be seen as aninterim step until STEP is spread offin industry. Furthermore, the results in developing,using and testing the dataJormatof this interim step will lead to a practical experience whichinputs to the developmentof aSTEP Application Protocol.

The exchange of theRP relevantgeometry informationofparts between systemsis based on 3D data, e.g. neutral datamodels/(lnitiaLGraphics Exchange Specification(IOES), interfaces ofthe 'Verband derdeutschen Automobilindustrie' - VDA-FS, VDA-IS-/2/, etc.) or facet formats developed for RP (STL, CFL, etc.). Slice routines or allthese formats. But the commonly used facet format is STL. Nearly all related CAD vendorsoffer an STL interface. The disadvantages of the STL syntax are sufficiently discussed.Nevertheless, there is no need for a new 'de facto' facet standardfor RP. Titne and energy tospread off a new facet format very high compared with the benefits. Other facet formats,fitting some applications better should be used individually, but should not be used or spreadoff as an exchange format. Experience made in RP or other areas using facet formats (e.g.Virtual Reality and Finite Element Methods) should be included in the developments ofSTEP. Nevertheless, benefits of individual facet solutions in special application will bedemonstrated later by the example of reverse engineering.

An essential point of the interim data handling is a 'de facto' standard for sliceinformation additional to STL! Direct CAD slice interfaces become even more important.Unfortunately, there is a wide range of variations in slice formats. Nearly every vendor usesits own slice format e.g. SLC, CLI, HPOL. The content of this different slice formats variesfrom pure geometry information up to machine specific data.

Figure 3 gives an overview of the data handling using one common facet format andone common slice format. The most important benefit of the two 'de facto' standards will be,that users will have the choice to select the data handling (direct slice interface or slicing ofSTL) depending on which one fits their problems best.

12

SLFigure 3 : Data handling in RP

~ /

~t computing of facet file out of slice

t slice of facet fileIIIC: ::J

ne1utralsice

-.,format.../ generation +control code~

~ ~MJS

CAD

directslice interface

The opinions on direct slicing and on slicing of facets break up in a wide range,dependant on experiences, habits and strategical aims. From a mathematical point of view,there is no difference. The accuracy depends on the quality of the software and not on thechoice. The mathematical expenditure to compute a facet format and slice it afterwards is ingeneral equal to the direct slicing. In practise, there are a lot of differences in quality,ergonomics, time and costs dependant on the available software and on the specialapplication. In our opinion, the users should be able to decide themselves. But in fact, theusers mostly do not have the choice, because there is a lack of a 'de facto' slice standard anddue to this, a lack of software tools supporting direct slicing.

Improved facet description for reverse engineering

Reverse engineering plays an important role in Rapid Product Development.Independant of the individual sensor the result of the digitizing process is point data.Software for automatic detection and computation of regUlar geometries and freeformsurfaces based on a spline description is complex and difficult to realize. The currentavailable solutions are limited in their abilities. As shown in figure 3, an alternative to link

13

digitizing to RP is to compute a facet presentation of the geometry. A problem is, that a parthas to be measured from different views. The result is a number of records with overlappingareas. The hardware installation and software at IPA enables to compute a facet descriptionof the measured areas. Contraction of the point data set and smoothing of the surface is alsopossible. Furthermore, a software tool to handle, to change and to close the surfaceaccording to different applications was developed.

The objective of this tool will become clear with the following example of application."The hilt of a butcher knife is often designed by manualmanufacturing of a physical model.A property of this model is, that one side of the hilt is not the exact reflecting surface of theother. So, the first step to produce a prototype is to determine the better side and to digitize itfrom selected directions. The point data set leads to a facet file describing the better side ofthe hilt and due to the measurement process some areas of the other side. To achieve a closeddescription of the whole hilt the facet file has to be limited by a plane and the mirror imagehas to be computed."

Data files in reverse engineering. do not seldom have millions •.of facets. Therefore,speed is an important request for the software tools. The STL format turned out to be lesssuitable to that issue. Access time to selected information is too slow. Therefore we use anown format designed for this demand. Similar to CFL we have a list with the values of allpoints. The triangles are described using the numbers of the edge points. In addition, we sortthe triangles related to the software tools used for special applications. Optional filesdescribing the sorted stock enable very quick algorithms to handle the facet information. Ifwe use our facet format, sort it according to the application and do the mapping to an STLfile at the end, it turned out that this way is much faster than handling the STL file.

Neutral slice format

Slices are the lowest common denominator among RP systems. For several reasons theusers request for a neutral slice format becomes even higher. The choice between differentRP systems in product development increases. Users want to apply the same software fordifferent RP systems as well as they prefer direct slice. Some benefits of a 'de facto' sliceformat are:

Q Software developers have free access to the syntax. The free access simplifies thedevelopment of software tools improving the data handling. For the users the choicebetween different software tools would raise.

Q CAD vendors are surely more willing to develop a slice interface, if there is onecommon format.

Q In medical application the output of tomographs is slice information. A common directlink to nearly all RP systems would be available.

Q Corrections of failures in the geometry description are easier to make in some cases.Q In developing new RP systems there is no need to develop an individual slicer.

Admittedly, there are a lot of applications where the use of slice information hasdisadvantages. One is that it is not usable to exchange geometry information for LOM,because the slice thickness varies online during the process. Furthermore, there aredisadvantages in manipulating the part e.g. rotating of sliced parts. But once again, it shouldnot replace a facet format but it should supplement it. The decision which one to use for thedifferent applications is task of the users.

14

Examinations to such a slice fonnat were done in a European project. The fonnatrequirements and the results have been published /3/. The fonnat Layer Exchange ASCIIFonnat (LEAF) was developed but has not been ryalized. It has an object oriented structurewhere instances of an object of given classes can be created. The only 2D primitivesupported by LEAF are polylines. The physical representation of LEAF uses keywordsdetennining the following syntax.

I...--S_Ii_c_e_fi_le_---'~__p_a_rt_ ___'~I__l_a~y_e_r_.....J~I 2D primitives IFigure 4 : Object class tree ofLEAF 131

Further work was done in another European project. Recently CLI, the slice fonnat ofEOS, has been improved. The new version should be suitable for most RP systems. Systemdependant infonnation can be presented in the header section. Unfortunately, the latestversion is not distributed until now, so a final comment is not possible.

One task of the slice fonnat at IPA is to run MJS as well as a Laser Modeling SystemofFockele & Schwarze. The second task is to get experience for the development of the slicepart of an RP STEP Application Protocol. Therefore, an object class tree (figure 4) as it isused in LEAF is the basic structure, too. The slice file should describe the product'sgeometry. However, the geometry of the RP part often deviates from this according to aspecial application. For investment casting the stereolithography part is produced with theQuickCast build style or only the shell of the part is produced. With MJS it is often ofadvantage to use special fill strategies due to the following sinter process. E.g. the fonnatmust be capable to contain this infonnation.

Regular geometries are often sliced in direct slice processes. Therefore, the slice resultis a summary of regular 2D primitives, e.g. lines, arcs, circles and ellipses. These primitivesshould be handled as primitives and not as a sum of polygon lines due to storage space.

Good experience was maid with a structure that allows a direct link between 'inner'borders to the surrounding'outer' border (figure 5). The computation of NC code turned outbe easier and faster.

outborder 1

outborder 2

oI

••~. inborders

\

outborder 3

start outborder 1line .line .start inborder 1

circle ...start inborder 2

line ...

start inborder 3circle

start outborder 2

start outborder 3circle ...start inborder 1

circle ...

Figure 5: Slice and diagrammatic structure of slice 2D primitives

15

Efforts to bring the of CLI, LEAF and those made during the design ofIPA's fonnat together and to coordinate the activities were started. Different questions aboutcontents and syntax of the fonnat have to be discussed. But it will surely be a compromisebetween demand to an early available format and the demand to a 'perfect' fonnat.Nevertheless, long term aim is

Conclusion

advantage of the MJS-process is the high flexibility of materials, e.g. highmelting metals andceramics,and.the simplicity of the apparatus. The future developmentswill improve accuracy and.the .number ofmaterials.

The applications of theRP technologies are still increasing. Easy and choice ofRPsystems and available software tools fitting to the special equipment and application ofthe users is required. Theacceptanceofa 'de facto' slice standard by all RP vendors wouldprobably improve the present situation in data handling. Both a facet and a slice descriptionshould be part of a STEP based data handling in RP

References

111 Kennicott,P.R.: An Application Reference Model for Layered Manufacturing. In:Proceedings of the IMS International Conference on Rapid Product Development,January 31 February 2, 1994, Fraunhofer-Institute for Manufacturing Engineeringand Automation, Stuttgart, pp 451-473

121 Renz, W.: Initiative des VDA zur Verbesserung des CAD Datenaustausches. In:Proceedings of the 8th International Exhibition for Computer Aided Technologies,May 5-8, 1992, Stuttgart, pp. 19-28

131 Dolenc, A.: Software Tools for Rapid Prototyping Technologies in Manufacturing.Acta Polytechnica Scandinavia, Mathematics and Computer Science Series No. 62,Helsinki, 1993

16